Compare
Self-hosted product analytics.
Product analytics records what people do in your product and answers who did what. Self-hosted means the software, and every event it records, lives on a server you run. It has a cost as well as a benefit.
Product analytics, not web analytics
Web analytics counts visits: how many people came, from where, to which pages. There are many good self-hosted tools for that. Product analytics asks a different question: which people did what, in what order, and did they come back. That cannot be answered without deciding that an anonymous visitor from two weeks ago and the account that signed up on Tuesday are one person.
So the thing to check in any tool in this category is identity: how it ties anonymous activity to a known account, what happens when two people share a device, and whether two ids that turn out to be one person can be merged.
Who offers it
Claims about other products below link each vendor's own pages, checked on 22 September 2026.
- PostHog can be self-hosted. Its documentation calls the self-hosted build a “hobby deploy” that is “unlikely to scale past a couple 100ks events without significant effort in scaling”, and says self-hosted installs get the features PostHog offers free, with every paid feature in PostHog Cloud. Lyraflow and PostHog compared.
- Mixpanel is hosted. Its documentation describes US, EU and India data residency, and no self-hosted option. Lyraflow as a Mixpanel alternative.
- Amplitude is hosted, from AWS data centres in the US and the EU. Its documentation describes no self-hosted option. Lyraflow as an Amplitude alternative.
- Lyraflow is self-hosted only. There is no hosted version.
What self-hosting gets you
Product analytics means recording, in detail, what your customers do. Self-hosting keeps that record on infrastructure you already control. There is no data-processing agreement to sign, because there is no processor. There is no per-event bill to forecast, because nobody is metering you. And there is nobody to ask for an export, because you already hold the database.
What it costs you
You run a server, and everything that comes with one is your job: HTTPS, backups, upgrades, disk space, and noticing when it is down. Nobody else is on call for it. If your team does not want that work, a hosted tool is the better choice.
What running Lyraflow involves
You need Docker and Docker Compose. ./install.sh generates the
passwords, starts three containers (the app, ClickHouse for events and Postgres for
identity) and prints a one-time admin password. Given a domain, it also starts
Caddy and serves HTTPS itself. Getting started
is four steps and about five minutes, most of it waiting for Docker.
After that, the operations are written down. One script backs up both databases together and another restores them. Events older than a project's retention setting, 13 months by default, are dropped on a schedule. A project can be given a monthly event cap. Upgrading is a backup, a rebuild and a restart, and migrations run on boot.
What Lyraflow does not do yet
There is no session replay, no feature flags, no experiments and no surveys. There is no path analysis, no alerting, no scheduled export and no email digest. A funnel and a retention grid cannot be broken down by a dimension the way a trend can. There is one admin account and no roles. Ingest never fills country, region or city yet, because there is no GeoIP lookup (#12); the countries in the demo are generated seed data. There is no importer, and the ingest API clamps timestamps to within 24 hours, so history from another tool cannot be brought across. The shipped install runs everything on one server, there is no published container image yet, and there is no published minimum machine or measured event volume.
Lyraflow is fair-code under the Sustainable Use License, not open source: you may self-host, use and modify it for your own business, but not sell it as a hosted service. Its first commit was on 1 August 2026, it is built by one person, and as far as we know nobody outside the project runs it in production yet.
Try it before installing it
The live demo is a running Lyraflow full of generated people and events, rebuilt every hour. The documentation covers every screen and endpoint, and says what each screen does not do yet.